Merge pull request #147056 from mweinelt/smartctl-exporter

nixos/modules/services/monitoring/prometheus/exporters.nix

@@ -61,6 +61,7 @@ let
     "rtl_433"
     "script"
     "snmp"
+    "smartctl"
     "smokeping"
     "sql"
     "surfboard"

nixos/modules/services/monitoring/prometheus/exporters/smartctl.nix

@@ -0,0 +1,64 @@
+{ config, lib, pkgs, options }:
+
+with lib;
+
+let
+  cfg = config.services.prometheus.exporters.smartctl;
+  format = pkgs.formats.yaml {};
+  configFile = format.generate "smartctl-exporter.yml" {
+    smartctl_exporter = {
+      bind_to = "${cfg.listenAddress}:${toString cfg.port}";
+      url_path = "/metrics";
+      smartctl_location = "${pkgs.smartmontools}/bin/smartctl";
+      collect_not_more_than_period = cfg.maxInterval;
+      devices = cfg.devices;
+    };
+  };
+in {
+  port = 9633;
+
+  extraOpts = {
+    devices = mkOption {
+      type = types.listOf types.str;
+      default = [];
+      example = literalExpression ''
+        [ "/dev/sda", "/dev/nvme0n1" ];
+      '';
+      description = ''
+        Paths to disks that will be monitored.
+      '';
+    };
+    maxInterval = mkOption {
+      type = types.str;
+      default = "60s";
+      example = "2m";
+      description = ''
+        Interval that limits how often a disk can be queried.
+      '';
+    };
+  };
+
+  serviceOpts = {
+    serviceConfig = {
+      AmbientCapabilities = [
+        "CAP_SYS_ADMIN"
+      ];
+      CapabilityBoundingSet = [
+        "CAP_SYS_ADMIN"
+      ];
+      DevicePolicy = "closed";
+      DeviceAllow = lib.mkForce cfg.devices;
+      ExecStart = ''
+        ${pkgs.prometheus-smartctl-exporter}/bin/smartctl_exporter -config ${configFile}
+      '';
+      PrivateDevices = lib.mkForce false;
+      ProtectProc = "invisible";
+      ProcSubset = "pid";
+      SupplementaryGroups = [ "disk" ];
+      SystemCallFilter = [
+        "@system-service"
+        "~@privileged @resources"
+      ];
+    };
+  };
+}

nixos/tests/prometheus-exporters.nix

@@ -1018,6 +1018,25 @@ let
       '';
     };
 
+    smartctl = {
+      exporterConfig = {
+        enable = true;
+        devices = [
+          "/dev/vda"
+        ];
+      };
+      exporterTest = ''
+        wait_for_unit("prometheus-smartctl-exporter.service")
+        wait_for_open_port("9633")
+        wait_until_succeeds(
+          "curl -sSf 'localhost:9633/metrics'"
+        )
+        wait_until_succeeds(
+            'journalctl -eu prometheus-smartctl-exporter.service -o cat | grep "/dev/vda: Unable to detect device type"'
+        )
+      '';
+    };
+
     smokeping = {
       exporterConfig = {
         enable = true;

pkgs/servers/monitoring/prometheus/smartctl-exporter/0001-Return-the-cached-value-if-it-s-not-time-to-scan-aga.patch

@@ -0,0 +1,51 @@
+From e81b06df67b1d42ef915615fafa0b56ef956673b Mon Sep 17 00:00:00 2001
+From: Andreas Fuchs <asf@boinkor.net>
+Date: Thu, 11 Feb 2021 17:30:44 -0500
+Subject: [PATCH] Return the cached value if it's not time to scan again yet
+
+This should ensure that if we have a valid value cached (which ought
+to be every time after the first scan), we return it as metrics.
+
+This fixes the crashes that would happen if queries happened earlier
+than the re-scan interval allowed.
+
+Address review feedback: Shorten the time-to-scan logic
+
+We can express this in a single if statement, so it takes fewer lines
+to do the "should we check again" check.
+---
+ readjson.go | 11 ++---------
+ 1 file changed, 2 insertions(+), 9 deletions(-)
+
+diff --git a/readjson.go b/readjson.go
+index da35448..c9996fd 100644
+--- a/readjson.go
++++ b/readjson.go
+@@ -78,14 +78,7 @@ func readData(device string) (gjson.Result, error) {
+ 
+ 	if _, err := os.Stat(device); err == nil {
+ 		cacheValue, cacheOk := jsonCache[device]
+-		timeToScan := false
+-		if cacheOk {
+-			timeToScan = time.Now().After(cacheValue.LastCollect.Add(options.SMARTctl.CollectPeriodDuration))
+-		} else {
+-			timeToScan = true
+-		}
+-
+-		if timeToScan {
++		if !cacheOk || time.Now().After(cacheValue.LastCollect.Add(options.SMARTctl.CollectPeriodDuration)) {
+ 			json, ok := readSMARTctl(device)
+ 			if ok {
+ 				jsonCache[device] = JSONCache{JSON: json, LastCollect: time.Now()}
+@@ -93,7 +86,7 @@ func readData(device string) (gjson.Result, error) {
+ 			}
+ 			return gjson.Parse(DEFAULT_EMPTY_JSON), fmt.Errorf("smartctl returned bad data for device %s", device)
+ 		}
+-		return gjson.Parse(DEFAULT_EMPTY_JSON), fmt.Errorf("Too early collect called for device %s", device)
++		return cacheValue.JSON, nil
+ 	}
+ 	return gjson.Parse(DEFAULT_EMPTY_JSON), fmt.Errorf("Device %s unavialable", device)
+ }
+-- 
+2.33.1
+

pkgs/servers/monitoring/prometheus/smartctl-exporter/default.nix

@@ -0,0 +1,41 @@
+{ lib
+, fetchFromGitHub
+, fetchpatch
+, buildGoModule
+}:
+
+buildGoModule rec {
+  pname = "smartctl_exporter";
+  version = "unstable-2020-11-14";
+
+  src = fetchFromGitHub {
+    owner = "prometheus-community";
+    repo = pname;
+    rev = "e27581d56ad80340fb076d3ce22cef337ed76679";
+    sha256 = "sha256-iWaFDjVLBIAA9zGe0utbuvmEdA3R5lge0iCh3j2JfE8=";
+  };
+
+  patches = [
+    # Fixes out of range panic (https://github.com/prometheus-community/smartctl_exporter/issues/19)
+    (fetchpatch {
+      url = "https://github.com/prometheus-community/smartctl_exporter/commit/15575301a8e2fe5802a8c066c6fa9765d50b8cfa.patch";
+      sha256 = "sha256-HLUrGXNz3uKpuQBUgQBSw6EGbGl23hQnimTGl64M5bQ=";
+    })
+    # Fix validation on empty smartctl response (https://github.com/prometheus-community/smartctl_exporter/pull/31)
+    (fetchpatch {
+      url = "https://github.com/prometheus-community/smartctl_exporter/commit/744b4e5f6a46e029d31d5aa46642e85f429c2cfa.patch";
+      sha256 = "sha256-MgLtYR1SpM6XrZQQ3AgQRmNF3OnaBCqXMJRV9BOzKPc=";
+    })
+    # Fixes missing metrics if outside of query interval (https://github.com/prometheus-community/smartctl_exporter/pull/18)
+    ./0001-Return-the-cached-value-if-it-s-not-time-to-scan-aga.patch
+  ];
+
+  vendorSha256 = "1xhrzkfm2p20k7prgdfax4408g4qpa4wbxigmcmfz7kjg2zi88ld";
+
+  meta = with lib; {
+    description = "Export smartctl statistics for Prometheus";
+    homepage = "https://github.com/prometheus-community/smartctl_exporter";
+    license = licenses.lgpl3;
+    maintainers = with maintainers; [ hexa ];
+  };
+}

pkgs/top-level/all-packages.nix

@@ -21405,6 +21405,7 @@ with pkgs;
   prometheus-rabbitmq-exporter = callPackage ../servers/monitoring/prometheus/rabbitmq-exporter.nix { };
   prometheus-rtl_433-exporter = callPackage ../servers/monitoring/prometheus/rtl_433-exporter.nix { };
   prometheus-script-exporter = callPackage ../servers/monitoring/prometheus/script-exporter.nix { };
+  prometheus-smartctl-exporter = callPackage ../servers/monitoring/prometheus/smartctl-exporter { };
   prometheus-smokeping-prober = callPackage ../servers/monitoring/prometheus/smokeping-prober.nix { };
   prometheus-snmp-exporter = callPackage ../servers/monitoring/prometheus/snmp-exporter.nix { };
   prometheus-statsd-exporter = callPackage ../servers/monitoring/prometheus/statsd-exporter.nix { };